森林植生季節性光譜特性之研究

Study of Seasonal Changes of Spectral Reflectance of Forest Vegetation

林金樹(Chin-Su Lin)

《台灣林業科學》 14卷3期 (1999/09) Pp. 289-305

https://doi.org/10.7075/TJFS.199909.0289

摘要

本文利用大地衛星TM多譜資料探討季節性變化對森林植群光譜反射特性之影響，希望瞭解不同植群的光譜特徵異質性與季節變化之關係，決定利用多譜資料檢測森林植群的最適時期與資料處理之邏輯及方法。研究結果發現，季節性變化對植物的光譜反射量有很大的影響，而且不同植群對不同波段的光譜反射量，因季節推移所引起的變化量並不一致；但若以光譜反射特性曲線的型態看，植物與裸地反射可見光和近紅外光光譜的基本特性是沒有改變的，而且裸地的光譜反射特性曲線迥異於植物，證明在任一季節利用遙測多譜資料檢測林區非法的開墾行為，是絕對可行的。春季時期是植物生長季節的開始，植物的冠層結構與葉部總面積因生長所生之變化極微；春季時期的太陽位置比較接近天頂，衛星影像因地形效應所引起的陰影面積比較少，可檢測的影像面積比率比較高；不同季節影像之組類分離度與分類準確度顯示，春季影像中不同植物的光譜特徵異質性最大，組類可辨率是四個季節影像中最高的。綜合這些條件可以證明，春季是利用遙測多譜資料檢測森林植群資訊的最佳時期。依地形與太陽高度可以模擬得到日照反射率影像，純化反射率的方法可以縮小地物每一細部組類光譜反射值的變異程度，並且降低地形效應造成同一地物之光譜反射值變異量擴大的影響。利用遙測技術監測森林植群時可以利用反射率純化法提升地物細部組類的光譜分離度。

關鍵字

森林植群 ； 遙測 ； 光譜反射特性 ； 季節性變異 ； 日照反射率純化法

並列摘要

This paper uses Landsat TM data to study differences of spectral reflectance and the spectral heterogeneity of different vegetation types caused by seasonal changes. It aims to find the best imaging season and sound data processing logic for monitoring forest vegetation. Results revealed that there is a significant influence of vegetative reflectance produced by seasonal change, yet the quantity changing in each TM band is not consistent. In spite of this fact, the basic reflection properties of visible and near infrared wavelength Light are almost identical and distinguishable in a year's duration for both vegetation and bare land if viewed from the pattern of spectral reflectance curves. This implies that it is possible to detect illegal development in a forest with remotely sensed data. Photosynthesis can directly and indirectly affect the visible and near infrared reflectance of trees, such as conifers, broadleaf trees, and bamboo; while background materials can also enhance the spectral reflectance of grass as a result of the photosynthesis process. In spring, the beginning of the plant growth season, physiological processes rarely change the total leaf area and canopy structure; furthermore the solar elevation is high enough to decrease the shadow area in the spring image. These conditions indicate that the spectral characteristics of plants are most stable in springtime. Further considering the best classes divergence and classification accuracy of the spring image revealed in this study, one can conclude that spring is the most suitable season for detecting forest vegetation with TM data. In the presence of the relief effect in an image, the spectral reflectance deviation of an object always expands. The light reflectance purification method separates the reflectance of an object into several classes, and hence can minimize the data deviation of each class. Results also revealed that the purification method is better than the relative proportion transformation method in reducing terrain effects in this study; and hence one can use this logic to process multispectral scanner data to upgrade the accuracy of classification in forest vegetation monitoring.

並列關鍵字

forest vegetation ； remote sensing ； spectral reflectance characteristics ； seasonal change ； and purified reflectance analysis